Someone has asked whether customers are happy, and the honest answer right now is a guess built from complaints and a few sales calls. A customer satisfaction survey replaces the guess with a number you can track and a set of comments you can read. This article, part of our guide to customer surveys that produce decisions, covers what a CSAT survey is, how to calculate the score, how it differs from NPS and CES, and what to do with the results.
Key takeaways
- A customer satisfaction survey collects customers’ ratings of a specific experience, usually one closed rating question followed by an open question asking why.
- CSAT is the percentage of respondents choosing 4 or 5 on a five-point scale, so 130 of 200 responses gives a score of 65%.
- CSAT measures satisfaction with a moment, NPS measures the overall relationship, and CES measures the effort a task required, so the three answer different questions.
- Send transactional surveys within minutes or a day of the event, milestone surveys at defined journey points, and relationship surveys once or twice a year.
- Act on results by splitting CSAT by segment, finding the drivers, reading the comments and re-measuring, because a healthy 65% overall can hide 40% in one group.
What is a customer satisfaction survey?
A customer satisfaction survey collects customers’ ratings of a specific experience: a purchase, a support conversation, an onboarding call, a delivery, or the product as a whole. The core is one closed question (“How satisfied were you with…?”) answered on a scale, often followed by an open question asking why.
Two things make it useful. The closed question gives you a score you can compare across time, segments and teams. The open question gives you the reason behind the score, which is where the fixes come from.
How do you calculate a CSAT score?
The most common method counts the share of respondents who chose the top two points of a 5-point scale (1 = very dissatisfied, 5 = very satisfied):
CSAT (%) = (number of 4 and 5 responses ÷ total responses) × 100Worked example
Suppose 200 customers rate their satisfaction with a support interaction. These numbers are illustrative.
| Rating | Label | Responses |
|---|---|---|
| 5 | Very satisfied | 70 |
| 4 | Satisfied | 60 |
| 3 | Neutral | 40 |
| 2 | Dissatisfied | 20 |
| 1 | Very dissatisfied | 10 |
| Total | 200 |
- Add the satisfied responses: 70 + 60 = 130.
- Divide by total responses: 130 ÷ 200 = 0.65.
- Multiply by 100: CSAT = 65%.
Some teams report the average rating instead. Here that would be (5×70 + 4×60 + 3×40 + 2×20 + 1×10) ÷ 200 = 760 ÷ 200 = 3.8. Either method works, but pick one and keep it. Mixing them across waves makes trends meaningless.
How is CSAT different from NPS and CES?
All three are single-question survey metrics, but they measure different things. CSAT measures a moment, NPS measures the relationship, and CES measures friction. For the full method behind the second one, see our guide to what Net Promoter Score is.
| CSAT | NPS | CES | |
|---|---|---|---|
| Question | How satisfied were you with…? | How likely are you to recommend us? | How easy was it to get your issue handled? |
| Typical scale | 1 to 5 | 0 to 10 | 1 to 7 (agree or disagree) |
| Score | % answering 4 or 5 | % promoters minus % detractors | Average, or % agreeing |
| Measures | Satisfaction with an interaction or product | Overall loyalty | Effort required of the customer |
| Best used | Right after an event | Periodically, as a relationship check | After a service or self-service task |
Used together they cover more ground than any one alone. A customer can be satisfied with a single support call (high CSAT) and still be unlikely to recommend you (low NPS) because of price or product gaps.
What questions should a customer satisfaction survey ask?
Keep the survey short and tied to one decision. A good starting set:
- Overall satisfaction: “Overall, how satisfied are you with [product or experience]?” (1 to 5)
- Driver ratings: “How satisfied are you with [speed of delivery / product quality / helpfulness of support]?” (1 to 5, one question per driver)
- Open comment: “What is the main reason for your rating?”
- Improvement prompt: “What one thing would most improve your experience?”
- Segment fields: plan type, tenure, region, so you can compare groups. Where possible, take these from your own records rather than asking.
Write each question about one thing only. “How satisfied are you with our price and quality?” cannot be answered honestly by someone who likes one and not the other. The open questions are worth the extra reading; our piece on the benefits and challenges of open-ended survey questions explains how to handle them.
When should you send a customer satisfaction survey?
Match the timing to what you are measuring.
- Transactional: send within minutes or a day of the event (a support case closed, an order delivered), while the memory is fresh.
- Milestone: send at a defined point in the journey, such as 30 days after onboarding or after a renewal.
- Relationship: send on a regular schedule, for example once or twice a year, to the same type of customer each time.
Whatever you choose, keep the cadence and the wording consistent. If you survey the same customer after every interaction, response and honesty both fall. Track your survey response rate so you know whether the people answering resemble the people you serve.
How do you act on customer satisfaction results?
A score on its own changes nothing. Turn it into decisions with four steps.
- Break it down. Split CSAT by segment, product, channel and time period. A healthy overall 65% can hide a 40% in one group.
- Find the drivers. Compare the driver ratings with overall satisfaction to see which aspects move the score most. To find the satisfaction and loyalty drivers behind each segment, you can start from the same data.
- Read the comments. Group open-text answers into themes and count them. The theme named most often among low scorers is your first candidate for a fix.
- Close the loop and re-measure. Follow up with dissatisfied customers where you can, make one or two changes, and run the same survey again to see whether the number moved.
Check that a change is real before you celebrate it. A shift of a few points between two small samples can be noise, so look at the sample size and a significance test before reporting a trend.
If you analyze survey results in Halo Reports, you can build crosstabs and banner tables that split CSAT by segment, apply weighting and significance testing, and share the results as charts or interactive report pages. See Halo Reports for what it covers. For a continuous view between waves, mTab for customer experience teams covers satisfaction drivers, journey friction and emerging issues.
What are common mistakes with customer satisfaction surveys?
- Surveying only happy or only unhappy customers. If the survey is sent after good events, or reaches only people who complain, the score will not describe your customer base.
- Changing the scale between waves. Moving from a 5-point to a 7-point scale, or from “top two boxes” to the average, breaks comparison with earlier results.
- Leading or double-barreled questions. “How satisfied were you with our excellent service?” nudges the answer, and two-topic questions blur it.
- Ignoring social pressure. Customers rate more kindly when they think a named person will read their answer. Our article on why people lie on customer surveys covers how to reduce that.
- Collecting scores and never acting. A survey that produces no visible change teaches customers that answering is pointless.
What should you do next?
Pick one moment in the customer journey that matters, such as the close of a support case. Write a two-question survey: a 1 to 5 satisfaction rating and an open “why”. Send it for a month, calculate CSAT with the formula above, read every comment, and choose one fix. Then run the same survey again.